[etherlab-dev] [etherlab-users] Lost slaves after device restart

[Stefan Agner]

Hello, 

I found my Problem: I requested a master while Ethernet was not connected by 
using ecrt_request_master(0). I immediately call ecrt_master_deactivate (this 
didn't helped because the functions exits very early because the master was not 
activated) and ecrt_release_master.

The Function ec_master_enter_operation_phase sets "allow_scan" to 0. As I said, 
the master gets never activated, it just leaves its state by using 
ec_master_leave_operation_phase. But allow_scan don't get resetted then, the 
master never ever rescans the bus...!


I solved the problem by adding two lines (master->allow_scan = 
1/master->allow_config = 1) to the function ec_master_clear_config, which gets 
always called:

void ec_master_clear_config(
        ec_master_t *master /**< EtherCAT master. */
        )
{
    down(&master->master_sem);
    ec_master_clear_domains(master);
    ec_master_clear_slave_configs(master);

    master->allow_scan = 1;
    master->allow_config = 1;

    up(&master->master_sem);
}

Maybe somebody should review this before adding to the repository... It solved 
my problem.

Bye
Stefan



----- Original Message ----
From: Stefan Agner <stefan.agner at yahoo.com>
To: etherlab-users at etherlab.org
Sent: Tue, May 10, 2011 1:48:43 PM
Subject: [etherlab-users] Lost slaves after device restart

Hi,

I use 1.5 trunk version of the Etherlab EtherCAT master on a Debian Lenny Linux 
5.0. I use the generic network driver (using default a device driven by a 
default e1000e driver) and PREEMPT_RT. 

Normally all works fine: I start Linux, Etherlab EtherCAT Master, and check the 
list of slaves:
# ethercat slaves
0  0:0  OP  +  EK1101 EtherCAT-Koppler (2A E-Bus, ID-Switch)
1  0:1  OP  +  EL6692 EtherCAT Bridge-Klemme (Primär)

I now start my application, which requests the master. The application runs 
fine.

When I shut down the EtherCAT device, slaves disappears as expcted:
# ifconfigeth0 down
# ethercat slaves
# 

Next I activate the device again, but slaves don't show up...
# ifconfigeth0 up
# ethercat slaves
# 

When I stop my application (release the master) the slaves still don't appear. 
Even rescan doesn't help, it looks like user space missed this new state (device 

up)... 


In dmesg everything looks ok:
Disabling:
[16091.763659] EtherCAT WARNING 0: 3 datagrams TIMED OUT!
[16091.783764] EtherCAT 0: Link state changed to DOWN.
[16091.787793] EtherCAT 0: 0 slave(s) responding.

Enabling:
[16105.749575] e1000e 0000:02:00.0: irq 28 for MSI/MSI-X
[16105.805136] e1000e 0000:02:00.0: irq 28 for MSI/MSI-X
[16105.806045] ADDRCONF(NETDEV_UP): ethercat0: link is not ready
[16107.390075] e1000e: ethercat0 NIC Link is Up 100 Mbps Full Duplex, Flow 
Control: None
[16107.390082] 0000:02:00.0: ethercat0: 10/100 speed: disabling TSO
[16107.390521] ADDRCONF(NETDEV_CHANGE): ethercat0: link becomes ready
[16107.390966] EtherCAT 0: Link state changed to UP.
[16107.395015] EtherCAT 0: 2 slave(s) responding.
[16107.395022] EtherCAT 0: Slave states: PREOP.

How can I make sure state and the slaves are tracked correctly in user space?

Thanks for any hints
Stefan

_______________________________________________
etherlab-users mailing list
etherlab-users at etherlab.org
http://lists.etherlab.org/mailman/listinfo/etherlab-users